At roughly 425 parts per million, today’s atmospheric carbon dioxide concentration sits nearly 80 percent above the level that, according to new modeling, triggers runaway ice loss in Antarctica. A study published in May 2026 in Nature Geoscience finds that the Antarctic Ice Sheet is far more sensitive to CO₂ shifts than the models guiding current sea-level planning. The research identifies a sharp, nonlinear threshold near 240 parts per million by volume. Below that line, small CO₂ changes produced outsized swings in ice volume over the past three million years. Above it, the ice sheet appeared deceptively stable. The implication for a world already well past that threshold: Antarctica could shed mass faster, and less predictably, than any mainstream forecast has assumed.
The core finding and why it diverges from older projections
The paper, titled “Increased sensitivity of the Antarctic Ice Sheet to decreasing CO₂ across the Mid-Pleistocene Transition,” is built on a new generation of ice-sheet simulations driven by a continuous, three-million-year climate reconstruction called CESM-3Ma. Developed at the IBS Center for Climate Physics, that dataset reconstructs how temperature, precipitation, and ocean heat delivery evolved together over deep time, rather than forcing an ice-sheet model with isolated snapshots of past climates.
When the researchers fed that continuous record into their ice-sheet model (labeled PSUIM-3Ma), the results broke sharply from older work. Previous landmark studies, including the widely cited 2016 Nature paper by Robert DeConto and David Pollard that introduced hydrofracturing and ice-cliff collapse as potential accelerants, relied on shorter or less continuous climate forcing. Those earlier runs captured important physics but missed the threshold behavior that emerges only when the model is allowed to evolve through millions of years of glacial cycles.
The new simulations show that during the Mid-Pleistocene Transition, roughly 1.25 million to 700,000 years ago, when Earth’s glacial cycles shifted from 41,000-year to 100,000-year spacing, the Antarctic Ice Sheet was already reacting sharply to CO₂ changes at concentrations well below today’s levels. If the ice sheet was that reactive at 240 ppmv, the researchers argue, it will respond at least as strongly to the higher and still-rising concentrations of the present era.
What the geology beneath the ice reveals
The threshold behavior is not just a modeling artifact. It has a physical explanation rooted in the shape of the ground beneath the ice. Subglacial topography datasets, including the Bedmap2 compilation and higher-resolution BedMachine surveys, show that in many sectors of Antarctica the bedrock slopes downward as it extends inland from the coast. Grounding lines, the points where ice lifts off the seabed and begins to float, often sit on narrow ridges that act as physical brakes.
Once warm ocean water pushes a grounding line past one of those stabilizing ridges and onto a reverse slope, retreat can accelerate under its own momentum. Each meter of retreat exposes a thicker cross-section of ice to warm water, which drives further melting, which drives further retreat. The 2026 study couples that well-established physics to its three-million-year forcing history and finds that CO₂ levels act as a master control on how close the system sits to those tipping points.
What remains uncertain
Several important pieces are still missing. The full time-series mass-loss rates at the 240 ppmv threshold have not been released in a form that outside groups can independently replicate field by field. The CESM-3Ma dataset is accessible through standard data interfaces, but the companion PSUIM-3Ma output remains limited to an archived package rather than a fully open release.
Direct observational validation is another gap. Satellite missions like GRACE-FO and ICESat-2 track modern ice-mass changes in near real time, yet no published study has compared the modeled grounding-line migration rates from the three-million-year runs against those satellite records. That comparison would test whether the model’s heightened sensitivity to CO₂ translates into realistic retreat speeds under current conditions.
There is also the question of reversibility. A 2020 study in Nature by Garbe and colleagues demonstrated that once the Antarctic Ice Sheet retreats past certain thresholds, it does not simply regrow when forcing reverses, a phenomenon known as hysteresis. The 2026 paper builds on that concept, but a quantitative comparison of hysteresis width between the two sets of experiments has not been published. Without it, the degree to which the new runs predict more or less reversibility remains an open question.
And the study’s primary focus is the deep past, not the 21st century. Translating its sensitivity findings into specific future timelines will require blending these long transient simulations with higher-resolution regional models and contemporary observations.
What this means for sea-level planning
Current planning benchmarks, including the IPCC’s Sixth Assessment Report, project roughly 0.3 to 1.0 meters of global sea-level rise by 2100 under moderate-to-high emissions scenarios, with Antarctic contributions treated as a major but uncertain component. The new study does not replace those numbers with a single revised figure. What it does is widen the plausible upper bound by showing that the ice sheet can shift between states more abruptly than the linear or slowly accelerating trajectories most planning frameworks assume.
For cities like Miami, Jakarta, Shanghai, and London, which are already investing billions in flood defenses and managed retreat, the practical consequence is a need to stress-test infrastructure against scenarios that include abrupt jumps in ice loss, not just smooth curves. Sediment cores and ancient shoreline markers already suggest that Antarctica contributed several meters of sea-level rise during warmer intervals of the Pleistocene, such as Marine Isotope Stage 11 and the mid-Pliocene warm period. The new three-million-year runs show that such large swings can be triggered by CO₂ shifts that are small compared with the difference between pre-industrial levels (about 280 ppmv) and today (about 425 ppmv). That alignment between independent lines of evidence strengthens the case that the ice sheet is not the sluggish, near-inert system once imagined.
The research does not claim that Antarctic collapse is imminent or unavoidable. It argues that the ice sheet is more tightly coupled to atmospheric CO₂ than legacy models captured, which makes emissions trajectories over the next few decades especially consequential. Pathways that limit peak CO₂ and stabilize temperatures could reduce the likelihood of pushing key grounding lines past their points of no return, even if some additional sea-level rise is already locked in.
A narrower margin than planned for
As additional PSUIM-3Ma data fields are released and more research groups test the results against satellite observations, the scientific picture will sharpen. For now, the strongest reading of the evidence is that Antarctica’s ice is both more fragile and more dynamic than the models behind current coastal plans suggested. The window for keeping the ice sheet’s most extreme responses off the table is narrower than most communities have built into their defenses, and the CO₂ levels that matter most are not hypothetical futures. They are the concentrations already in the atmosphere.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
